Why assume open-weight model providers will stay open forever?

Atlan_ · reddit · 2026-10-08

A Reddit discussion challenges the assumption that open-weight providers will remain open. The logic: "open when behind, closed when ahead." Models from DeepSeek, Mistral and other open labs have found spots on the efficient frontier and keep consuming more token volume; with questionable profitability, they have every incentive to charge licensing fees like closed vendors — staying free is money left on the table. The author speculates NVIDIA might keep models open to suppress LLM prices and boost demand, but sees no reason for other labs not to charge.

Original post →

More from Venture

Venture channel →